Revised manuscript, CHERD-D-13-00961-
,
Yang Zhang
1
, Yanbin Jiang
1
, Duanke Zhang
1
, Yu Qian
1 and Xue Z. Wang
1,2*
1 School of Chemistry and Chemical Engineering, South China University of Technology,
Guangzhou 510640, China
2
Institute of Particle Science and Engineering, School of Process, Environmental and
Materials Engineering, University of Leeds, Leeds, LS2 9JT, UK
*
To whom correspondence should be addressed.
E-mail: xuezhongwang@scut.edu.cn or x.z.wang@leeds.ac.uk

Abstract
The production of

-Artemether, 3, 12-epoxy- 12h-pyrano (4, 3- j)- 1, 2- benzodioxepin,
C
16
H
26
O
5
), an important active pharmaceutical ingredient for the treatment of malaria, uses crystallization as the purification step. This paper investigates the Metastable Zone Width
(MSZW), nucleation and crystal growth kinetics of anti-solvent crystallization for

artemether in ethanol + water system where water is the anti-solvent. Using a turbidity probe, the effect of process variables on MSZW was investigated, including initial the mass ratio of the solute to solvent, temperature, introduction rate as well as introduction location of the anti-solvent. The prediction model for MSZW was calibrated and proved reliable for this system. For anti-solvent crystallization of

-artemether in ethanol + water system, the chord length distribution data was collected using Focused Beam Reflectance Measurement, based on which the nucleation and crystal growth parameters were estimated using the method of moment. The crystal growth mechanism of

-artemether was also studied using a hot stage and imaging system and was found to follow a layer-by-layer growth mechanism.
Keywords :

-Artemether; FBRM; Crystallization; Kinetics; Mathematical modelling;
Thermodynamics
1

1. INTRODUCTION

-Artemether (3, 12-epoxy-12h-pyrano (4,3-j)- 1,2-benzodioxepin, C
16
H
26
O
5
) is probably the most effective active pharmaceutical ingredient (API) recommended by the
World Health Organization (WHO, 2006) for treatment of malaria, a global killer of around
800,000 people a year and is second only to tuberculosis in its impact on world health. The parasitic disease is present in 90 countries and infects one in 10 of the world's population - mainly people living in Africa, India, Brazil, Sri Lanka, Vietnam, Colombia and the Solomon
Islands. For this important API, crystallization is often used as a key purification process for quality control (Zhang et al., 2009) in its production, but there has been little data about its crystal growth kinetics in the public domain.
As an important industrial separation and purification process, crystallization process control has been the subject of active research (Nagy and Braatz, 2012; Nagy et al., 2008;
Nowee et al., 2008b; Yu et al., 2007), with the focus being on crystal size distribution (CSD) as well as polymorph and morphology control of the solid active pharmaceutical ingredients(APIs) or active ingredients for fine chemicals such as agrochemicals. These properties impact not only on the applied performance of crystalline products, but also on downstream processes such as filtering, drying, transportation and storage (Fujiwara et al.,
2005; Nagy et al., 2013). CSD and crystal polymorph and morphology are determined by the nucleation and crystal growth kinetics (Barrett et al., 2010; Feng and Berglund, 2002;
Lewiner et al., 2001; Yang and Nagy, 2013). Understanding of the kinetic models is, therefore, the prerequisite for achieving effective crystallization process control. However, to our best knowledge, there have been no reports concerning the fundamental crystallization kinetics of

-artemether. In this paper, the crystallization kinetics of

-artemether are investigated using on-line process analytical technology (PAT). Prediction models for metastable zone width
(MSZW) and a kinetics equation of

-artemether are developed and calibrated. In addition,
2

the crystal growth mechanism is also investigated.
Traditionally, off-line sampling was used for calibration of kinetic models, but the complicated sampling process is not only time consuming, but also means that the quality of samples cannot be assured, leading to inconsistency. More recent recommendations are the use of advanced on-line PAT with high quality sensor probes. Concentration, particle size, and polymorphism can be measured in real time and batch to batch variations can be minimized or avoided resulting in improved efficiency of energy and materials (Borissova et al., 2008;
Calderon De Anda et al., 2005; O'Sullivan et al., 2003; Zhang et al., 2011).
Focused Beam Reflectance Measurement (FBRM) is used in the study. An immersed
FBRM probe is used to obtain the chord length distribution (CLD) of particles in the range from 1nm to 1000nm in real time based on a backscattering mechanism. FBRM has been used by previous researchers to help characterize the metastable zone width (MSZW) and polymorphic transformation (Barrett and Glennon, 2002). Sullivan and Glennon (O'Sullivan and Glennon, 2005) utilized FBRM for research into cooling crystallization, including identifying polymorphic transformation. Barrett and Glennon (Barrett and Glennon, 2002) developed an on-line measurement method for solubility curves and MSZW using FBRM.
FBRM was also utilized as the tool to characterize both the nucleation and crystal growth kinetics. Czapla et al. (Czapla et al., 2010) developed a FBRM model to quantify preferential crystallization of DL-threonine, the kinetic parameters of preferential crystallization were estimated based on FBRM CLD data. Trifkovic et al.(Trifkovic et al., 2008) researched the crystallization kinetics using FBRM and CLD was directly used for parameters estimation of crystal growth kinetics using the method of moment, which was regarded as a paradigm for kinetics measurement.
3

2. Experimental
2.1. Materials

-Artemether was obtained from the Pidi Pharmaceutical Co. in China with a stated purity > 99.5% by mass and an experimental melting point of 87.7 o
C. All other reagents were of analytical grade and were purchased from Guangdong Guanghua Sci. Tech. Co. Ltd, China.
All chemicals were used as obtained. Redistilled water was used in all experiments.
2.2.
Experimental apparatus and procedure for measurement of MSZW
The experimental apparatus for measuring MSZW data is shown in Figure 1 and all
MSZW measurements were performed in a 200 mL jacketed crystallizer. The turbidmeter
(Mettler Toledo, Trb8300) was used for obtaining the turbidity information, and a thermostat
(Ningbo Tianheng Instrument Inc., THD 0506) was used for controlling the temperature of the crystallizer. A metering pump (Shanghai Jingke Instrument, HL-2B) was used to manipulate the flow rate of water added to the crystallizer.
MSZW was measured using the following the procedure: prepared

-artemether ethanol solution was added into the crystallizer first, then the temperature of the crystallizer was adjusted to set point. After making sure all the

-artemether was dissolved, water was added and the on-line turbidity data was recorded simultaneously.
Single factor experiments were carried out to investigate the effect of process variables on MSZW, including the mass ratio of solute to solvent, temperatures, adding rates of antisolvent and adding locations of anti-solvent. The selected experimental conditions were as follows, i.e., temperature 298.15 K , 303.15K and 308.15 K, adding rate for water 2 g·min
-1
, 5 g·min -1 , 8 g·min -1 , 12 g·min -1 and 15 g·min -1 , stirring speed 200 rpm and 400 rpm,

mass ratios of

-artemether to ethanol 5/70 and 5/90, respectively. Also two adding locations for water were chosen, one nears the shaft, and the other nears the wall of the crystallizer.
4

2.3.
Experimental apparatus and procedure for measurement of kinetics
The experimental apparatus used for kinetic measurement is shown in Figure 2. All kinetic experiments were conducted in a 500 mL semi-batch crystallizer. A FBRM (Mettler
Toledo, Redmond, WA) was used for collecting the CLD data. A metering pump (Shanghai
Jingke Co., HL-2B) was used to manipulate the flow rate of water added to the crystallizer.
The procedure for kinetic measurement is described as follows: prepared

-artemether solution containing

-artemether and ethanol was added into the crystallizer, then the stirring speed was set. The temperature of the initial clear solution was kept at 5 degrees above the saturated temperature for 30 minutes, making sure all chemicals were dissolved in the solvent
(Yang and Nagy, 2013). The measured particle number of initial solution shown in the FBRM was less than 150. Then the temperature of the prepared

-artemether solution was lower to the set point to start the experiment, simultaneously the anti-solvent was added into the crystallizer for generation of supersaturation. All CLD data was scanned once every 20 s during the process. After particles were detected in the solution, the slurry was sampled and then filtered using a vacuum filter (1
 m mesh). Solution concentration and slurry density were calculated using the gravimetric method.
All kinetic experiments described in this work were carried out using unseeded crystallization. Experimental conditions for kinetics experiments were shown in Table 1, which includes eight different runs. The adding rate of the anti-solvent was selected as 1.25 mL · min
-1
. Experimental temperatures were set at 288.15 K , 303.15 K , 308.15 K , 313.15 K , and 318.15K, and stirring speeds ( N r
) were 250 rpm, 300 rpm, 350 rpm and 450 rpm, respectively.
2.4.
Apparatus and procedure for identification of the crystal growth mechanism
An on-line image analysis system consisting of a Linkam hot stage (Linkam Scientific
Instrument, LTS 350), a microscope (Olympus, TH4-200) and a high resolution CCD camera
5

(Sony, XC-55) were used for identification of the crystal growth mechanism in this study.
Before starting the experiment, a prepared supersaturated solution of

-artemether in ethanol
+ water was added into the cell of the hot stage, then the temperature programme of the
Linkam hot stage was set, crystal growth in the cell of the hot stage was observed using the microscope and pictures of the crystal were taken every 0.5 s using the CCD camera and recorded simultaneously.
3. Results and Discussion
3.1. Results of MSZW

A max
is the maximum excessive anti-solvent composition defined by

A max
A A sat
, where A is the anti-solvent composition calculated at the outset of nucleation( t ) detected by on
-line turbidity probe, A sat
is the anti-solvent composition of solution saturated with the solute
(Kubota, 2008). In this study, A sat
was calculated from the literature (Zhang et al., 2009).
MSZW of the

-artemether solution at different experimental conditions are given in Table 2.
The process variables investigated included temperature, stirring speed, adding location and adding rate of the anti-solvent, and the mass ratio of solute to solvent.
3.1.1. Effect of temperature on MSZW
MSZW of the

-artemether in ethanol + water system at different temperatures is plotted in Figure 3. Figure 3 shows that MSZW increased significantly as crystallization temperature changed from 298.15 K to 303.15
K , and decreased as crystallization temperature changed from 303.15 K to 308.15 K . Actually, MSZW is affected by corresponding solubility data and the Gibbs free energy barrier. As the solubility increased with the increasing of temperature, it became more difficult to form nuclei in the solution. However, the change of Gibbs free energy will help formation of nuclei when increasing temperature.
3.1.2. Effect of initial mass ratio of solute to solvent on MSZW
6

The effect of initial mass ratio of solute to solvent on MSZW is given in Figure 4. Figure
4 shows that the MSZW of

-artemether in ethanol + water system becomes narrow when the mass ratio of solute to solvent increases, which indicates that the lower initial concentration will delay the formation of nuclei.
3.1.3. Effect of the addition location of anti-solvent on MSZW
The effect of addition location on the experimental MSZW of

-artemether in ethanol + water system is shown in Figure 5. Two adding locations were investigated, one near the shaft and the other close to the wall of the crystallizer. Figure 5 indicates that the MSZW of adding location for anti-solvent near the wall is wider than that of the adding location near the shaft, probably due to the difference in mixing efficiency. The mixing efficiency of the addition location near the shaft was considered better. Therefore the preferred addition location was chosen to be near the shaft.
3.1.4. Effect of stirring speed on MSZW
The effect of stirring speed on MSZW is shown in Figure 6. Two stirring speeds were investigated, 200 rpm and 400 rpm, respectively. The result shows that stirring speed is an important impact factor for MSZW, when stirring speed increases, the MSZW becomes wider.
It indicated that local supersaturation near the introduction point of anti-solvent is quite high when stirring is not sufficient.
3.1.5. Effect of adding rate on MSZW
It can be seen from Figures 3-6 that MSZW becomes wider when increasing the adding rate of anti-solvent. This phenomenon is similar to the effect of cooling rate on MSZW during cooling crystallization.
3.1.6. Development of a Prediction Model for MSZW
The nucleation rate can be expressed as Equation (1) (Nývlt, 1968):
𝐵
0
= 𝐾
𝐵
γ
∆C max
(1)
7


C max
was supersaturation in the solution, K
B
was the nucleation rate constant, γ was the constant coefficient.
The nucleation rate can also be expressed based on the generation rate of supersaturation, as shown in Equation (2):
B

0 qb (2) where b =dA/dt, which is the rate of addition of anti-solvent. q is deposited solid per unit solution, and q
 
( dC eq
/ dA ) p , where
 
R /[1

(

1)]
2 , and R is the ratio of molecular weights of hydrate. As there was no solvate of

-artemether, ξ equals 0 in this study.
For anti-solvent crystallization, maximum supersaturation can be expressed with the maximum allowable anti-solvent concentration, as shown in Equation (3) (O'Grady et al.,
2007):

C max

 dC eq
/ dA


A max
(3)
Equation (3) can be transformed into Equation (4):

 dC eq
/ dA
 p b

K
B
 dC eq
/ dA


A max
Equation (4) can be abbreviated to Equation (5): ln

A max

K met

O met ln
 dC eq
/ dA


Q met ln b
(4)
(5)
During anti-solvent crystallization, stirring speed was considered as an important impact factor. In this work, effect of stirring speed ( N r
) on the maximum allowable anti-solvent concentration was investigated and shown in Equation (6), N r
was stirring speed, K met
, O met
,
P met
, and Q met
were calibration constant in Equation (6).

A max

K met

 dC eq
/ dA

O met
N r
P met
 b
Q met (6)
The model for prediction of MSZW of

-artemether in ethanol + water system based on the measured MSZW data was calibrated using the least squares method in MATLAB software, which is given as:
8


A max
 
5 
 dC / dA eq
 
0.7666
0.0930
0.3095
N b r
(7)
R
2 
0 .
88
Comparison between experimental MSZW data and the calculated MSZW value is given in Figure 7. The result shows that the deviation between calculated MSZW values and experimental MSZW data is acceptable.
3.2. Results of Kinetics
To investigate the kinetics for the anti-solvent crystallization of

-artemether, eight selected runs were conducted according to the given experimental conditions shown in Table
1, and typical experimental CLD results (run S5 in Table 1) of

-artemether are given in
Figures 8 and 9. In the early stage of anti-solvent crystallization, no particles were detected in the solution, however, the formed nuclei are detected by FBRM when anti-solvent was added after 3 mins, which indicated that nucleation occurred in the crystallizer. From Figure 8, CLD became wider with time elapsed from 0 to 19.0 mins, this indicated crystal particles grow lager and second nucleation occurs in the solution.
The population balance equation is the most useful method for development of the kinetics equation as shown in Equation (8) (Myerson, 2001):
 n
 t


 

L

Q o
V n

Q j
V n i

( B
 
D

) (8) where G is the crystal growth rate, B

is the birth function, D

is the death function, L is the crystal size , V is system volume, Q is the flow rate across the boundary, t is time, and n is the population balance function, which can be calculated using Equation(9): lim

L

0

N

L
 dN dL
 n
(9)
In Equation (9), N is the total number of particles. When aggregate and breakage can be
9

ignored, and crystal growth obeys McCabe’s Law, Equation (9) can be abbreviated to
Equation (10):
 n
 t



L
 n d (ln V ) dt

0 (10)
In this work, the kinetics equation is calibrated using the method of moment, and the j th moment of population balance function is given in Equation (11) (Togkalidou et al., 2004):
 j
  
0 nL j dL j =0, 1, 2, 3 (11)
The j th
moment
μ j of population balance function n by differentiation with respect to time t are as follows(Nagy et al., 2008): d

0 dt
 d

0
 n ( L , t ) dL dt
 dN dt

B d

1 dt
 d
 
0 n ( L , t ) LdL dt

G

0 d

2 dt
 d

0
 n ( L , t ) L 2 dL dt

2 G

1 d

3 dt
 d

0
 n ( L , t ) L 3 dL dt

3 G

2
(12)
(13)
(14)
(15)
The j th
moment of experimental CLD data obtained from FBRM can be calculated using
Equation (16) (Trifkovic et al., 2009):
 j
 
0
 j

FBRMfinal channel i


1 r j ( i ave,i
, ) (16) where r and N i
are the chord length and particle number respectively. r ave, k
is the mean of chord length between two consecutive channels ( i and i +1), k is the discrete time counter.
The experimental nucleation rate B exp, j and crystal growth rate G exp,j
can be calculated from the 0 th
moment and the 1 st
moment based on CLD data (Trifkovic et al., 2009), as shown in Equations(17) and (18) (Sheikhzadeh et al., 2008):
10

B exp, j

1
M j
 dN dt

1
M initial

M antisolven t , j


N
 t j j
(17)
G exp, j



 d
 
0 dt


 j



 d ( L
1 , 0 dt
)


 j





L
1 , 0
 t


 j
(18)
During the anti-solvent crystallization process, the nucleation rate and crystal growth rate are functions of temperature, mass fraction of anti-solvent and supersaturation, as shown in
Equations (19) to(23) (Nowee et al., 2008a):
𝐵 𝑝𝑟𝑒𝑑,𝑗
= 𝐾 𝑏
∆𝐶 𝑗 𝛾 𝛾 = 𝛾
1 𝑚 𝑤,𝑘
+ 𝛾
2
G

K g
 c g j
K g

K g 0 exp


E g


 RT j
 g
 g
1 m w , k
 g
2
(19)
(20)
(21)
(22)
(23) where B pred,j and G pred,j
are the nucleation rate and crystal growth rate at j time, respectively.
K b
and K g are the nucleation rate constant and the growth rate constant, respectively. m w, k is the weight fraction of solvent in solvent mixture, T j
is temperature at j time, and
γ
1
,
γ
2
, g
1
and g
2 are constant coefficients.
Based on the experimental CLD data obtained from eight selected runs using FBRM, the equations of nucleation rate and crystal growth rate of

-artemether in ethanol + water system are calibrated and given in Equations (24) and (25):
B
  



 5
RT




C
0.38
m

0.54
M t

2.05
N r
1.77
R
2 
0 .
77
G

4.13exp




666.31
RT




C
0.49
m

0.77
11
(24)
(25)

R 2

0 .
74
From Equations (24) and (25), kinetic models for

-artemether in ethanol + water developed in this work are of moderate accuracy, and the Rsquared values in Equations (24) and (25) are lower than 0.8. Two reasons are ascribed to the moderate R -squared value. Firstly, the measured CLD and number of particles are affected by the location of the FBRM probe in the solution, and the FBRM probe for different experimental runs is hard to keep in the same position. Secondly, FBRM is restricted for poor conversion technology from CLD data to actual CSD of crystals. Although many studies have been reported in this field, complicated mathematical models for conversion from CLD to CSD were affected by the crystal shape and agglomeration behaviour of the particles, thus the mathematical conversion model from CLD to PSD remains a topic for research (Tadayyon and Rohani, 1998).
3.3. Results of the Crystal Growth Mechanism
In this study, the crystal growth mechanism of

-artemether was investigated using an imaging system. Figure 10 shows a typical picture for crystal growth of

-artemether, which is observed using an on-line image analysis system. It indicates that the crystal growth of

artemether in an ethanol + water system obeys a layer by layer mechanism. According to this mechanism, crystals will grow uninterrupted under a lower solute concentration.
4. Conclusions
The effects of stirring speed, introduction rate of anti-solvent, temperature, initial ratio of solute to solvent as well as introduction location of anti-solvent on the MSZW of

artemether crystallization in ethanol + water system were investigated using an on-line turbidity probe. The model for prediction of the MSZW of

-artemether in ethanol + water system based on the measured MSZW data is calibrated, which is given as follows:
12


A max

4.4796 10

5 
 dC / dA eq
 
0.7666
0.0930
0.3095
N b r
Comparison between experimental MSZW data and calculated MSZW value shows the deviation between the two is acceptable.
CLD for anti-solvent crystallization of

-artemether was monitored and measured using
FBRM, and the nucleation and growth equations of

-artemether in ethanol + water system are calibrated using the method of moment based on the CLD data. The equation of nucleation rate and crystal growth rate of

-artemether in ethanol + water system respectively is as follows:
B
  



 5
RT




C
0.38
m

0.54
M t

2.05
N r
1.77
G

4.13exp




666.31
RT




C
0.49
m

0.77
But the Rsquared values indicate that the two kinetic models developed in this work are of moderate accuracy. The on-line process analytical technology is of great potential for the quality control of the anti-solvent crystallization process, but accurate prediction models still need improved measuring methods and apparatus.
Also a layer-by-layer growth phenomenon is observed on the face of the

-artemether crystal, and the crystal growth mechanism of

-artemether in ethanol + water is confirmed as following the layer-by-layer growth mechanism.
Acknowledgements
The following financial support is acknowledged: National Natural Science Foundation of China (No. 21076084), Guangdong Provincial Science & Technology Project (No.
2011B050400013), the Fundamental Research Funds for the Central Universities (No.
2011ZZ0006), as well as the 1000 Talents Programme of China. Funding from the Chinese
Scholarship Council for this international collaboration is also acknowledged. The
13

1
,
2
K
B
K g
L
M m corresponding author would like to express his gratitude to the South China University of
Technology (SCUT) for funding his Laboratory of Pharmaceutical Manufacturing and
Crystallization Control at SCUT under the national 1000 Talent Scheme, as well as the UK
Engineering and Physical Sciences Research Council (EPSRC) for supporting research in crystallization process control (EP/H008012, EP/E045707, EP/C009541) at the University of
Leeds.
Nomenclature
Latin symbols
A
A max anti-solvent concentration, g anti-solvent
g solvent
-1 the maximum allowable anti-solvent concentration, g anti-solvent
g solvent
-1 nucleation rate, number g solvent
-1 min
-1
B
B
 b
C
C eq birth function adding rate of anti-solvent , g min
-1 concentration, g solute
g solvent
-1 equilibrium concentration, g solute
g solvent
-1
D

G g death function crystal growth rate, m min
-1 growth rate index constant coefficient [Equation (23)] nucleation constant [Equation (1)] growth constant [Equation (21)] crystal size, m mass of solvent, g weight fraction of solvent in solvent mixture
14

N i
N r n q
Q r r ave,k total number of particles stirring speed, min
-1 population balance function mass of crystalline substance deposited per unit of free solvent flow rate across the boundary chord length, m mean of chord length between two consecutive channels, ( i and i +1)
T
V

C max temperature, K volume, m
-3 maximum allowable supersaturation, g solute
g solvent
Greek symbols
 j j th
moment of population balance function
-1
γ
γ
1
, γ
2
Subscripts constant coefficient [Equation (1), Equation (20)] constant coefficient [Equation (20)] exp experimental pre prediction
References
Barrett, M., McNamara, M., Hao, H., Barrett, P., Glennon, B., 2010. Supersaturation tracking for the development, optimization and control of crystallization processes. Chemical
Engineering Research and Design 88, 1108-1119.
Barrett, P., Glennon, B., 2002. Characterizing the metastable zone width and solubility curve using Lasentec FBRM and PVM. Chemical Engineering Research and Design 80, 799-
15

805.
Borissova, A., Khan, S., Mahmud, T., Roberts, K.J., Andrews, J., Dallin, P., Chen, Z.-P.,
Morris, J., 2008. In situ measurement of solution concentration during the batch cooling crystallization of l-glutamic acid using ATR-FTIR spectroscopy coupled with chemometrics. Crystal Growth & Design 9, 692-706.
Calderon De Anda, J., Wang, X.Z., Lai, X., Roberts, K.J., 2005. Classifying organic crystals via in-process image analysis and the use of monitoring charts to follow polymorphic and morphological changes. Journal of Process Control 15, 785-797.
Czapla, F., Kail, N., Öncül, A., Lorenz, H., Briesen, H., Seidel-Morgenstern, A., 2010.
Application of a recent FBRM-probe model to quantify preferential crystallization of dlthreonine. Chemical Engineering Research and Design 88, 1494-1504.
Feng, L., Berglund, K.A., 2002. ATR-FTIR for determining optimal cooling curves for batch crystallization of succinic acid. Crystal Growth & Design 2, 449-452.
Fujiwara, M., Nagy, Z.K., Chew, J.W., Braatz, R.D., 2005. First-principles and direct design approaches for the control of pharmaceutical crystallization. Journal of Process Control
15, 493-504.
Kubota, N., 2008. An interpretation of the metastable zone width concerning primary nucleation in anti-solvent crystallization. Journal of Crystal Growth 310, 4647-4651.
 Lewiner, F., Klein, J.P., Puel, F., Févotte, G., 2001. On-line ATR FTIR measurement of supersaturation during solution crystallization processes. Calibration and applications on three solute/solvent systems. Chemical Engineering Science 56, 2069-2084.
Myerson, A.S., 2001. Handbook of Industrial Crystallization second edition ed. Elsevier
Butterworth-Heinemann Woburn,MA.
Nývlt, J., 1968. Kinetics of nucleation in solutions. Journal of Crystal Growth 3-4, 377-383.
Nagy, Z.K., Braatz, R.D., 2012. Advances and New Directions in Crystallization Control.
Annual Review of Chemical and Biomolecular Engineering 3, 55-75.
Nagy, Z.K., Fevotte, G., Kramer, H., Simon, L.L., 2013. Recent advances in the monitoring, modelling and control of crystallization systems. Chemical Engineering Research and
Design 91, 1903-1922.
Nagy, Z.K., Fujiwara, M., Woo, X.Y., Braatz, R.D., 2008. Determination of the kinetic parameters for the crystallization of paracetamol from water using metastable zone width experiments. Industrial & Engineering Chemistry Research 47, 1245-1252.
Nowee, S.M., Abbas, A., Romagnoli, J.A., 2008a. Antisolvent crystallization: Model identification, experimental validation and dynamic simulation. Chemical Engineering
Science 63, 5457-5467.
Nowee, S.M., Abbas, A., Romagnoli, J.A., 2008b. Model-based optimal strategies for controlling particle size in antisolvent crystallization operations. Crystal Growth &
Design 8, 2698-2706.
O'Grady, D., Barrett, M., Casey, E., Glennon, B., 2007. The effect of mixing on the metastable zone width and nucleation kinetics in the anti-solvent crystallization of benzoic acid.
Chemical Engineering Research & Design 85, 945-952.
O'Sullivan, B., Barrett, P., Hsiao, G., Carr, A., Glennon, B., 2003. In situ monitoring of polymorphic transitions. Organic Process Research & Development 7, 977-982.
16

O'Sullivan, B., Glennon, B., 2005. Application of in situ FBRM and ATR-FTIR to the monitoring of the polymorphic transformation of d-Mannitol. Organic Process Research
& Development 9, 884-889.
Sheikhzadeh, M., Trifkovic, M., Rohani, S., 2008. Real-time optimal control of an antisolvent isothermal semi-batch crystallization process. Chemical Engineering Science 63,
829-839.
Tadayyon, A., Rohani, S., 1998. Determination of particle size distribution by Par-Tec® 100: modeling and experimental results. Particle & Particle Systems Characterization 15,
127-135.
Togkalidou, T., Tung, H.H., Sun, Y., Andrews, A.T., Braatz, R.D., 2004. Parameter estimation and optimization of a loosely bound aggregating pharmaceutical crystallization using in situ infrared and laser backscattering measurements. Industrial & Engineering
Chemistry Research 43, 6168-6181.
Trifkovic, M., Sheikhzadeh, M., Rohani, S., 2008. Kinetics estimation and single and multiobjective optimization of a seeded, anti-solvent, isothermal batch crystallizer. Industrial
& Engineering Chemistry Research 47, 1586-1595.
Trifkovic, M., Sheikhzadeh, M., Rohani, S., 2009. Multivariable real-time optimal control of a cooling and antisolvent semibatch crystallization process. AIChE Journal 55, 2591-2602.
WHO, 2006. momograph on good agricultural and collection practices (GACP) for artemisia annua L. WHO press, Switerland.
Yang, Y., Nagy, Z.K., 2013. Model-Based Systematic Design and Analysis Approach for
Unseeded Combined Cooling and Antisolvent Crystallization (CCAC) Systems. Crystal
Growth & Design 14, 687-698.
Yu, Z.Q., Chew, J.W., Chow, P.S., Tan, R.B.H., 2007. Recent advances in crystallization control: anindustrial perspective. Chemical Engineering Research and Design 85, 893-
905.
Zhang, Y., Jiang, Y., Li, K., Qian, Y., 2009. Solubility of beta-artemether in methanol + water and ethanol + water from (288.85 to 331.95) K. Journal of Chemical & Engineering
Data 54, 1340-1342.
Zhang, Y., Jiang, Y., Zhang, D., Li, K., Qian, Y., 2011. On-line concentration measurement for anti-solvent crystallization of [beta]-artemether using UV-vis fiber spectroscopy. Journal of Crystal Growth 314, 185-189.
17

TABLES
Table 1
Experimental conditions for kinetics of

-artemether anti-solvent crystallization process
No.
S1
S2
S3
S4
S5
S6
S7
S8
T (K)
288.15
298.15
308.15
313.15
318.15
298.15
303.15
298.15
Table 2
Experimental MSZW data of

-artemether in ethanol+ water system
Nr (rpm)
250
250
350
300
350
350
300
450
T (K) Nr (rpm)
Adding location* in out in
Initial composition
5:70
5:70
5:70 b (g · min
-1
)
2 5 8 12 15
0.1791 0.3004 0.3267 0.3839 0.3946 298.15 200
298.15
303.15
200
200
0.2124 0.3124 0.3839 0.5153 0.4196
0.3199 0.4237 0.4889 0.5327 0.5784
308.15 200
298.15 200 in in
5:70
5:90
0.3069 0.3949 0.4583 0.5316 0.5544
0.2323 0.3694 0.4250 0.5361 0.5611
298.15 400
298.15 400 in out
5:70
5:70
0.2243 0.3171 0.3934 0.4296 0.4981
0.1429 0.2881 0.3496 0.4467 0.4731
* In is the adding location near shaft, and out is the adding location near wall of crystallizer
18

Figures
Figure1: Schematic diagram of the experimental apparatus for MSZW measurement
Figure2: Schematic diagrams of experimental apparatus for kinetics measurement
19

0.6
298.15 K 303.15 K 308.15 K
Calculated values
0.4
0.2
0 4 8 12
Feeding rate (mL · min -1 )
16 20
Figure3: MSZW of

-artemether in ethanol + water at different temperatures
0.8
5/70 5/90
Calculated values
0.6
0.4
0.2
0
0 4 8 12
Feeding rate (mL· min 1 1 ) )
16 20
Figure4: MSZW of

-artemether in ethanol + water at different initial solution concentrations
20

0.6
Shaft Wall
Calculated values
0.4
0.2
0
0 4 8 12
Feeding rate (mL • min 1 1 ) )
16 20
Figure5: MSZW of

-artemether in ethanol + water at different adding locations
0.6
200 rpm 400 rpm
Calculated values
0.4
0.2
0
0 4 8 12
Feeding rate (mL• min 1 1 ) )
16 20
Figure 6: MSZW of

-artemether in ethanol + water at different stirring speeds
21


Am
Figure7: Comparison of experimental MSZW data and calculated MSZW
0.0 min
3.3 min
4.6 min
8.0 min
14.3 min
19.0 min
10 100
Size of particle/(
 m)
Figure8: Variation of particle size distribution during anti-solvent crystallization
(No. S5)
22

Time/ ( h )
Figure 9: Typical plot of the total particle number and mean length chord length of particle with elapsed time (No. S5)
Figure 10: A hexagon spiral on the face of

-artemether crystal
23